@burkeholland Caveat: at it' current capabilities, it still requires an awful lot of my time and can't be left alone.
English
David Flanagan
5.7K posts
David Flanagan
@rawkode
Helping the world scale AI at @CoreWeave. Teaching developers Kubernetes on https://t.co/rOwpNaxDgk
Scotland, Europe Katılım Mart 2009
1.5K Takip Edilen13.3K Takipçiler
What is the max you would pay for AI per month before it would make more sense to write code by hand?
I have a number...
English
@sama I don’t need them smarter, faster, or cheaper; I need them consistent.
The fluctuation in model quality is far more than determinism. It’s when the US wakes up, it’s 1 week after model launch, 1 week before a launch.
I need to be able to trust it, and I can’t.
English
i keep thinking i want the models to be cheaper/faster more than i want them to be smarter
but it seems that just being smarter is still the most important thing
English
I don’t know what @OpenAI model we get when we use Codex via iOS, but I do know it’s not a good one.
It’s awful.
English
@harmicful The container runtime doesn't pull images or build them.
Those operations are also IO bound, not CPU bound.
English
Performance quiz: If we re-wrote Docker in Python, what impact would it have on container performance?
Most people say "slower." They're wrong.
👇
English
Codex is for everyone, for any task done with a computer
OpenAI@OpenAI
It's never been easier to do everyday work with Codex. Choose your role, connect the apps you use every day, and try suggested prompts. Codex helps with everything from research and planning to docs, slides, spreadsheets, and more.
English
@krzyzanowskim @ImLunaHey So you can play “which ones in sync?” all day? 😅
English
@lucperkins "hooks: onEnter: nix: xNix.#NixFlake"
cd dir; # cuenv activates the flake
English
I'm working on a direnv extension that essentially provides the functionality of `use flake` but bypassing Nix evaluation. So you `cd` into the directory and the Nix-built Bash script just runs. Basically nix-direnv but, like, the next level of magic.
English
I rank my AI models via ffs's per minute, and 5.5 has a lot less than any Opus model.
English
English
@dillon_mulroy @zebassembly @stolinski I’ve got a single user / single org forge ready to release to the world. I just need access to Artifacts to kill off the Rust binary.
Help? 😊
English
Do people think a "new github" will be able to handle this kind of massive dump of new code daily?
English
Codex with GPT 5.5 is as flawed as Claude Code with Sonet 4.7. Both are good at nailing implementations ad hoc. But the code they produce when left unattended is a disaster for long-term maintenance.
English
@kellabyte I feel this soo much. My CI/CD times are averaging 15-25 mins depending on the size of the project. Much sadness.
English
Considering moving HoloStore tooling away from Rust because god damn these compile times are getting soooooo bad. The more Rust I build the longer everything takes.
Keep HoloStore rust and tools something else?
Rust is not meant for automated coding. It just doesn’t shine.
English
Ummm, doesn’t Gemini already have access to a search engine?
Someone help me, please.
Exa@ExaAILabs
We're excited to partner with Google to offer Grounding With Exa inside of Gemini models! Using Exa's agent-first search, Gemini models can now access billions of websites, technical docs, papers, people, companies, and more. 10^18🤝10^100
English
my linkedin messages and invites is where things rot to death.
English
@stolinski GitHub was all about social and contributors, the next wave of software is personal.
Why would you want a contributor to your habit app when your agents build it to your spec?
Your projects need a home and that home is your forge; the software of Scott.
Very exciting times
English
@rawkode This is def what I was missing from the conversation.
English
@dillon_mulroy @jullerino We're the first people they're coming for ...
English
